
DaVoice
On-device SDKs for wake word, speaker recognition, VAD, speech-to-text and text-to-speech.

About DaVoice
DaVoice: On-Device Voice AI for Apps, Vehicles and Devices
DaVoice provides on-device voice AI SDKs covering wake word detection, voice activity detection, speaker recognition, speech-to-text and text-to-speech, available individually or as one integrated pipeline. The models run locally on the device, so voice features can work offline and audio does not need to be sent to cloud speech services.
Key Features
- Custom Wake Words: Custom wake word training in more than 50 languages and accents, delivered as on-device models that listen continuously in the background.
- Benchmarked Accuracy: DaVoice reports a 97% detection rate with fewer than 0.0001 false positives per hour in an independent third-party test covering 59 languages.
- Low Resource Use: Wake word engines use less than 2% CPU on modern mobile devices and use hardware acceleration (CoreML on iOS, NNAPI on Android) when available.
- Speaker Recognition: Creates a voice signature for each enrolled user and verifies it in real time, at the wake word stage, the transcription stage or both.
- Speaker-Aware Speech-to-Text: Streams partial and final transcripts and can limit transcription to the enrolled user's voice, ignoring other nearby speakers.
- Voice Activity Detection: Detects when someone is speaking, as part of the same on-device voice stack.
- On-Device Text-to-Speech: Generates spoken responses locally, with handling for audio sessions and the switch between listening and speaking.
- On-Device Voice Cloning: Clones a voice and lets it speak in any language, using a compact on-device model with emotional expression.
- Cross-Platform SDKs: Supports Android, iOS and macOS, React Native, Flutter, Java and Python, with GitHub repositories per platform.
Use Cases
- Automotive Assistants: In-cabin voice control and assistants activated by a wake word, such as the Volkswagen car assistant demo shown on the DaVoice site.
- Mobile Apps: Hands-free voice commands in iOS, Android and React Native apps, including apps used in noisy settings like fitness.
- Healthcare and Enterprise Capture: Local transcription and read-back for workflows where voice data should stay on the device.
- Smart Home and Wearables: Always-listening wake words and local speech output for embedded and battery-powered products.
Pricing
- POC / Academic / Non-Commercial: Free tier for developers and non-commercial use, with limited use, for early testing and integration.
- Enterprise: Custom quote based on platforms, number of devices or users and the technologies used. DaVoice states its pricing is at least 50% below top competitors.
- Startups can move from a free trial POC to a jump-start plan, and revenue-share or success-based models are available.
Getting Started
- Website: https://davoice.io
- Documentation: https://davoice.io/docs
- Pricing: https://davoice.io/pricing
DaVoice gives developers and device makers a local voice stack, from wake word to spoken reply, that runs on the device instead of in the cloud.

